credo-release
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted LaTeX source files and build logs, creating an attack surface for indirect prompt injection where malicious instructions could be embedded in the paper source text.
- Ingestion points: The scripts/check_tex.py script reads .tex files and follows LaTeX include/input directives to collect project files.
- Boundary markers: The skill does not provide specific instructions to the agent to ignore or delimit embedded instructions within the LaTeX content it reads.
- Capability inventory: The agent executes the provided Python script and interprets its output, which includes snippets of the paper title and abstract.
- Sanitization: The script strips LaTeX comments and uses regular expressions to find specific patterns, but it does not fully sanitize content before displaying snippets to the agent.
- [SAFE]: The bundled Python script uses only the standard library and performs no network operations or unauthorized file writes. File access is restricted to reading the paper source and logs as intended by the skill's release-checking purpose. The script correctly implements cycle detection to prevent infinite loops during file scanning.
Audit Metadata